============================
Have 200 MHz proton data on low concentration polysaccharide in 99.9% D2O
with no reference compound. Main peak in standard acquisition is -OH from
0.1 % H2O and polysaccharide -OH's.
Have data acquired over 120 hours with presaturation of water with decoupler
channeled through transmitter. AC-200 system with O1=O2 modification. This
data set has broad peak ( 4 ppm ) centered at 6.2 ppm. Assume that this is
due to water adsorbed onto probe coil supports and surface of nmr tube. Don't
have dry nitrogen supply to run spin turbine and VT unit. Relative humidity
in lab is 40-50%.
Have tried zeroing first data points to get rid of rapidly decaying solid
water signal. Get severe baseline roll and phase problems. Any suggestions
for removal of broad hump while maintaining integrity of rest of data?
============================
Summary of responses:
============================
You could try two things;
1. Get rid of the water and -OH resonances first by drying down the sample
a
couple of times in D2O before acquiring data
2. Treat the sample with chelex first as it sounds like you might also have
paramagnetic impurities. If you want details on how to do this, let me
know.
============================
============================
1. Try classical convolution difference. It works for most cases
adjusting 2 parameters, LB and weighting factor to be subtracted.
============================
============================
We have had a lot of success getting rid of PTFE backgrounds in 19F
spectra by zeroing first N data points then doing a backwards linear
prediction from N to point 1. My suggestion is to keep the number of
coefficients down in the LP but use as much of the FID a you can see
data in to do the prediction. Most processing and spectrometer software
can now do linear prediction, although the speed of it varies
dramatically Bruker's XWINNMR and WINNMR will both do a 50 coefficient
prediction of 24 points from 800 data points in 'no time' - other
software I have seen takes minutes!
============================
============================
Try running the VT unit warm (100 C) for a while, then run dry nitrogen for
spinner,VT and body air.
(You can use vent gas from your liquid nitrogen supply dewar for this)Do you
see this broad spectral component on an empty probe ? It could be an
artifact from the epoxy used in construction. Another posibility is
microphony or acoustic resonances in the probe?
Another approach is just wait a while before aquiring. The broad component
will relax faster.
There has also been some pusle sequences (about 10 years ago if I recall) in
solid state NMR literature to prevent acoustic resonances that used short
tailored 180 just before aquisition to null out the humps.I'll send along
the reference if I can find them.
Anothe technique you may want to consider is Rapid Scan Cross Correlation
NMR if your system supports it.
============================
============================
Find out the linewidth of the broad peak. Use line broadening
equaling the linewidth of the broad peak +/- 50-100 Hz . Subtract
the resulting spectrum from the original data. Process the
difference spectrum as usual. You have to experimenting on the first
LB to get the best result.
============================
============================
It could be that you are seeing grease in capacitors or some such thing; in
any case the 90 degree pulse for the impurity may be much longer than for
the sample. Someone once suggested doing a series of 4 90 pulses with
different phases; the normal signal gets a net 90 pulse and the signals for
which is is a much shorter pulse get returned to the Z axis. I never really
tried it though.
============================
============================
If you either zero the first several data points or left shift the fid to
remove the first points, you will perforce get frequency dependent phase
shifts and the concomitant baseline roll. The only way I know to correct
this type of problem is to use linear prediction to back calculate what the
first several data points 'should' have looked like. This works well on
my Varian VnmrX software, but I am not familiar with what might be available
for AC-200 data sets. I believe that Woody Conover's NUTS program may be
able to both import the AC data and do linear prediction on it.
============================
============================
Have you tried taking a spectrum of the empty probe and subtracting it from
the data? The drawback to this approach is if the probe changes tuning on
sample insertion.
============================
============================
Have you tried a short spin echo? The broad peak presumably has a short
T2, so you shouldn't need much of an echo delay to get rid of it.
============================
--------------------------------------------------------
Numare Spectralab, Inc.,
10 Summit Avenue, Unit 6, P.O. Box 96,
Berkeley Heights, NJ, 07922-0096,
Telephone 908-665-0066
Telefax 908-464-4192
Name: Lawrence Byrnes
E-mail: numare@cnj.digex.net (Lawrence Byrnes)
www: http://www.cnj.digex.net/~numare/
--------------------------------------------------------